Re: Case mapping of dotless lowercase letters

[Peter Kirk]

On 17/12/2003 11:29, Philippe Verdy wrote:

Peter Kirk wrote:
 

Conclusion: the right thing even for Turkish is to drop the dot on i 
before a circumflex.
   

I agree. The letter is rare enough to not create an exception here for
the removal of dot on the soft-dotted i followed by circumflex (which
is needed much more often in other languages that use 'Ã' and Ã'.
 

I'm not sure that rarity is a good argument, but we agree on the conclusion.

But by the same argument we would also want to drop 
the dot on dotless I.
   

I think you meant "But by the same argument we would also want to drop 
the dot on DOTTED I". I would not recommand it, this would make things
even worse and more complicated.

 

Indeed. Thank you for correcting my error.

If Turkish wants to remove the dot on "pseudo-dotted" I if followed by
a circumflex, the correct thing to do is then to use the ASCII dotless
I and add a circumflex or use its canonical equivalent
<LATIN CAPITAL LETTER I WITH CIRCUMFLEX>.
With the current specification, both of
        <LATIN CAPITAL LETTER I, COMBINING CIRCUMFLEX>, and
        <LATIN CAPITAL LETTER I WITH CIRCUMFLEX>
are canonical equivalents and must render the same, without the dot.
To display a dot, one can use one of the four canonical eqquivalents:
        <LATIN CAPITAL LETTER I WITH DOT ABOVE, COMBINING CIRCUMFLEX>
        <LATIN CAPITAL LETTER I WITH CIRCUMFLEX, COMBINING DOT ABOVE>
        <LATIN CAPITAL LETTER I, COMBINING DOT ABOVE, COMBINING CIRCUMFLEX>
        <LATIN CAPITAL LETTER I, COMBINING CIRCUMFLEX, COMBINING DOT ABOVE>
(one is the NFC form, another is the NFD form, two others are also
possible)
 

The problem which might arise is when someone applies Turkic casing 
operations to a Turkish text including i with circumflex (at least in 
NFD). The small version becomes <dotted I, circumflex>, which is wrong 
and looks wrong. The capital version becomes <dotless i, circumflex>, 
which is wrong but looks correct.

So the rules should be adjusted so that the normal casing rule, not the 
special Turkic one, applies when there is a circumflex. But perhaps not 
when there are other accents e.g. I would expect that an acute is 
sometimes used as a stress marker but it would then need to appear in 
addition to the dot on i and dotted I, cf. Lithuanian.

--
Peter Kirk
[EMAIL PROTECTED] (personal)
[EMAIL PROTECTED] (work)
http://www.qaya.org/